Stable Diffusion 2.0 - 4 New Models Overview

Stable Diffusion 2.0 delivers a number of big improvements and features versus the original V1 release, so let’s dive in and take a look at them.

New Text-to-Image Diffusion Models
Super-resolution Upscaler Diffusion Models
Depth-to-Image Diffusion Model
Updated Inpainting Diffusion Model

[parttimelarry]

You are quick! Loving this channel, thanks!

[ICHRISTER]

They did release 1.5 on hugginface, by RunwayML around a month ago.
the reason they switching to 2.0 is its because its all new, the decoder, the sampler, and even the res is now 768x768.

[alphacat]

So far the most en depth video I have seen on this thank you.

[bytexd_official]

Thank you very much for the overview! I had just heard about it and went straight to your channel.

[ruchirguitar]

This guy lives under a rock. 1.5 was released a month ago.

[erikschiegg]

I see trees of green, skies of blue, x-transformers, too
they really say seven-six-eight and I think to myself, what a wonderful prompt... da-di-da-du-uh!

[Aldraz]

So now we can make videos much more consistent with depth2img, that is pretty cool

[MarkWilder]

How do I install the new SD2 but keep my SD Automatic 1111 UI?

[RetroUpgrade]

short and concise just how i like my news :D, the AI looks amazing, to bad they going the route of NSFW removal somewhat limits prompt options, but it looks like a decent update over 1.5
keep up the good work

[jean-michelgilbert]

Now, all they need to do to compete with e-Diffi (Nvidia's new text-to-image model) is add a frozen T5 text-to-text model as a preprocessing step to improve its understanding of natural language.

[crispinotechgaming]

now we just need someone to finetune 2.0 to NAKED BODIES and licensed content

[jtjames]

Removing NSFW is what makes these models terrible at anatomy.

[twobob]

1.5 got hijacked. 2.0 provides a fresh start maybe

[RaineRed]

Artists and a lot of celebs stripped. 2.0 looks like SD without a soul.